o1, der erste Kernmitarbeiter, der kündigte, äußerte sich erstmals: KI ist der stärkste Hebel in der Geschichte, der Menschenkraft, Kapital und Code übertrifft.

Another core researcher who left OpenAI has spoken out!

Hyung Won Chung, who was just reported to have joined Meta, shared his profound thoughts on the future of AI: Artificial intelligence is becoming the most powerful leverage mechanism in history.

Hyung Won Chung and Jason Wei, who also left OpenAI, are long - term partners. Their collaboration can be traced back to the Google Brain era. They co - authored an important paper on model fine - tuning, "Scaling Instruction - Finetuned Language Models", as the first authors.

Jason Wei once praised him:

I was deeply impressed by Hyung Won Chung's ability to identify new paradigms and completely abandon any sunk costs. At the end of 2022, he realized the power of reinforcement learning and has been advocating it ever since.

During his time at OpenAI, Hyung Won Chung was a contributor to core projects such as o1, o1 - preview, and Deep Research.

However, in his speech, he didn't talk about specific technical details. Instead, he interpreted AI from a broader perspective - how it is completely changing the way we create value, from the individual level to the level of human civilization.

The speech took place two months ago, but he shared these views at the time when it was reported that he left OpenAI, especially how AI is transforming organizational forms and subverting the way of wealth creation. Perhaps it implies his unique understanding of the future development direction.

AI Becomes a Multiplier of Individual Abilities

Hyung Won Chung started his speech with a budding flower.

He pointed out that humans are naturally not good at perceiving slow changes that occur over the course of years. This "flaw" may make us seriously underestimate the magnitude of the changes brought about by AI.

Artificial intelligence may be the fastest - developing technology in history, but its development is not measured in minutes or hours, but in years or even decades.

When explaining what leverage is, Hyung Won Chung gave a precise definition: Through a certain mechanism, a small or even zero change in input can bring about a large or even huge change in output.

He believes that throughout history, the main ways of wealth creation have always relied on three types of leverage - human resources, capital, and code. Now, AI is becoming the fourth and most powerful form of leverage.

Human resource leverage is the oldest way of wealth creation.

Building the pyramids would have been an almost impossible task if relying solely on one person's strength. However, by hiring thousands of workers, an individual's input may remain the same or even decrease - the organizer doesn't need to carry the stones himself, but the output increases exponentially. The characteristic of this leverage mechanism is that it requires "permission". That is, you must get the consent of these workers, pay them salaries, and manage and coordinate them.

Even in today's society, human resources are still one of the main leverage mechanisms. Large enterprises expand their production scale and market influence by hiring tens of thousands of employees. In essence, they are still using human resource leverage.

Capital leverage became the main way of wealth creation in the 20th century.

Suppose you want to invest in a property worth $1 million, but you only have $200,000 in principal. By borrowing $800,000 from the bank, you use 20% of the funds to leverage 100% of the assets.

If the value of the property doubles to $2 million, your return rate may reach 400% or even 500%, far exceeding the 100% increase in the property's value itself. This is the power of capital leverage - using less of your own funds to control a larger - scale asset, thus amplifying the investment return.

The rise of capital leverage is inseparable from the development of the modern financial system. The emergence of financial tools such as banks, stock markets, and bond markets has made the flow and allocation of capital more efficient. Many huge fortunes in the 20th century were created through the ingenious use of capital leverage, especially in the financial industry. However, capital leverage is also a double - edged sword. The 2008 financial crisis is a typical example of systemic risks caused by the over - use of capital leverage.

Code and media represent a new type of leverage in the information age.

If you write an application, the first user gets a value of n from it. Then, when you are sleeping, the second user downloads and uses the program and also gets a value of n - your output has doubled without you doing any extra work.

This is possible because software has the characteristic of zero - marginal - cost replication. Once the code is written, the cost of serving one user is almost the same as serving 100 million users.

The principle of media leverage is similar. If he gives a lecture to 200 people and uploads the recording to YouTube, every additional person who watches it adds more value without him having to repeat the same work.

More importantly, this growth in value is largely unlimited. A video can be watched millions of times, and an application can be downloaded billions of times, with the creator only having to make one effort.

Recently, much of the value creation has benefited from this type of code leverage. The reason why technology companies in Silicon Valley can create huge value in a short period of time is precisely because they fully utilize the replicability of software. An application developed by a team of only a few dozen people can serve hundreds of millions of users globally, which was unimaginable in traditional industries.

However, he also reminded everyone that the advantages of leverage mechanisms will gradually diminish due to competition.

When a certain type of leverage is widely recognized and used, it becomes increasingly difficult to obtain excess returns through it. Now, it is much more difficult to start a pure software company or become a YouTuber than it was ten years ago because there are many competitors and the market is becoming saturated.

This is why we need to constantly look for new leverage mechanisms.

Artificial intelligence is precisely such a relatively new and emerging form of leverage.

Different from traditional leverage, AI can not only replicate and expand a specific human ability. More importantly, it can learn, reason, and create. This enables it to function as an independent leverage and also combine with other forms of leverage to produce a more powerful compound effect.

Agents Initiate Permissionless Compound Leverage

When talking about AI agents, Hyung Won Chung believes that this may be the most interesting research area in 2025 and beyond.

He explained that AI agents cleverly combine the advantages of two leverage mechanisms.

First, it can work for you like human resource leverage, just like hiring employees.

But more importantly, agents that exist in pure software form have the characteristics of code - if you want 10 agents to work together, you just need to make a copy. It's also effortless to add 2 more, and you don't need to get anyone's permission.

The power of this compound leverage has begun to show in the business world. You may have heard of some start - up companies with only 10 to 20 people but can generate hundreds of millions of dollars in revenue, which was almost unimaginable 10 years ago.

The secret weapon of these companies is the extensive use of AI agents to undertake tasks that originally required hundreds or even thousands of employees. Customer service, content creation, data analysis, code development, and other aspects can all be completed or assisted by agents.

A deeper - level impact is that this new type of leverage is changing the organizational form itself.

Traditional large - scale human collaboration has natural bottlenecks. Adding one person to a team of 100 people may not increase the output by 1%, or it may even decrease, because the additional coordination cost may offset the value brought by the additional human resource.

However, if each of these 100 people is equipped with an AI agent as an assistant, the overall output may double or even increase more, while the increase in coordination cost is negligible.

Hyung Won Chung described this change as "profound" because it is not just an improvement in efficiency but a fundamental change in production relations. When creating value no longer mainly depends on organizing and managing a large number of human resources but on how to design, train, and deploy AI agents, the power structure and wealth distribution methods of the entire society may change accordingly.

This permissionless compound leverage is opening up unprecedented possibilities for individuals and small teams.

In the last part of his speech, Hyung Won Chung raised his perspective to the level of the entire human civilization. He posed a profound question: If we consider all of humanity as a whole, what is our task or goal?

He believes that the most sustainable engine for growth and value creation is scientific progress, that is, the discovery of new knowledge.

Since the scientific revolution in the 17th century, wealth creation has shown a "hockey - stick" - shaped growth. However, in modern society, promoting scientific progress has become increasingly complex. The complexity of cutting - edge technologies may far exceed the ability of any single individual, and human intellectual growth has relatively stagnated and cannot keep up with the growth rate of scientific complexity.

Whenever we encounter such bottlenecks, we can always find ways to break through and create tools to clear the obstacles. This time, I think artificial intelligence is such a useful tool.

He especially emphasized the role of AI in connecting the knowledge of experts in different fields.

In the scientific community, experts are often scattered in different places and groups, and cross - professional cooperation is difficult. "Human knowledge is like peaks in a high - dimensional space, with large gaps between them, and artificial intelligence is like a shell that wraps these peaks and connects all professional knowledge."

At the end of the speech, Hyung Won Chung returned to the opening theme. He asked the audience to think: Considering AI as a new form of leverage, is it possible that you are underestimating the magnitude of this change?

Video Replay:

https://x.com/hwchung27/status/1945355238187393257

Reference Links:

[1]https://x.com/hwchung27/status/1945355238187393257

This article is from the WeChat official account "QbitAI". Author: Meng Chen. Republished by 36Kr with permission.